AbstractA model in the here discussed form is first of all a description of complex processes in nature. This starts with a description by words or graphics (conceptual model) and goes up to complex mathematical or numerical simulation models that run on supercomputers. The complexity of a model is determined by several factors. It can be simply limited by computational resources which involves the questions if enough computer power or storage media for the data output is available or if there are appropriate tools available to make sure the data can be evaluated in a reasonable way. Developing a model can be seen as an interactive and iterative process. You can use a model to reproduce and understand experimental findings whereas at the same time experimental results are used to refine models. The most important point to consider before you build or acquire a model is the purpose the model should serve. Conceptual models e.g., could be used to describe physical or chemical processes in the atmosphere be it in lectures or for scientific discussions of processes. Mathematical models allow a more profound investigation of physical and chemical processes. Ultimately, models that are used to simulate the dispersion of substances over e.g., the European continent or to predict ambient air concentrations and deposition rates in a high spatial and temporal resolution can be very comprehensive. They often require a lot of input variables (e.g., meteorology, emissions) and parameters (e.g., physical–chemical constants). Insufficient knowledge of these inputs may lead to erroneous results or misleading interpretation of the results. The repertory of numerical simulation models ranges from simple box models that can e.g., represent a closed system and contain only one substance in one compartment and thus require little computer power to elaborate three-dimensional grid models containing plenty of substances involved in a number of physical and chemical processes. A scientist who develops or applies a model has to balance the complexity and expense of the model with the available input variables and parameters, the available computational resources and the demanded precision of the model results (Jacobson 2005).